Item recommendation service

1. A computer-implemented method of selecting items to recommend to a user, the method comprising: identifying a plurality of items that, based on activity of the user, are deemed to be of interest to the user; for each of said items of interest, retrieving, from a pre-generated data structure that maps items to related items, a respective related items list; weighting the related items lists based on information reflective of the user's affinity for the corresponding items of interest, such that at least some of the related items lists are weighted differently than others, wherein weighting a related items list comprises weighting individual items on the related items lists, and wherein the amounts by which the items are weighted influences a likelihood that such items will be selected to recommend to the user, wherein weighting the related items lists comprises at least one of the following: (1) weighting a related items list by an amount that depends upon an amount of time that has transpired since the user performed an action that evidences the user's affinity for the corresponding item of interest, and (2) weighting a related items list by an amount that depends upon an explicit rating by the user of the corresponding item of interest; generating a recommendations list at least partly by combining the weighted related items lists, said recommendations lists including data values for respective items thereon, each data value being dependent upon (1) whether the respective item is on more than one of the related items lists, and (2) the weighting applied to the related items list or lists on which the respective item appears; and selecting, based at least partly on said data values, a subset of the items on the recommendations list to recommend to the user; said method performed by a computing system that comprises one or more physical servers.

2. The method of claim 1 , wherein each related items list comprises a set of relatedness values, each relatedness value corresponding to a respective item on the related items list and representing a degree to which that item is related to the corresponding item of interest, and wherein weighting the related items lists comprises multiplying each of the relatedness values in a list by the weighting amount.

3. The method of claim 1 , wherein weighting the related items lists comprises weighting a related items list by an amount that depends upon an amount of time that has transpired since the user performed an action that evidences the user's affinity for the corresponding item of interest.

4. The method of claim 3 , wherein said action is a purchase by the user of the corresponding item of interest.

5. The method of claim 1 , wherein weighting the related items lists comprises weighting a related items list by an amount that depends upon an explicit rating by the user of the corresponding item of interest.

6. The method of claim 1 , further comprising, in an offline processing mode, generating the data structure based at least partly on detected correlations between item interests of users.

7. The method of claim 1 , wherein the pre-generated data structure is based at least partly on content-based relationships between particular items.

8. The method of claim 1 , wherein identifying the plurality of items of interest consists of identifying items represented in an electronic shopping cart of the user.

9. The method of claim 1 , wherein identifying the plurality of items of interest comprises selecting at least some of the items of interest from a recorded item viewing history of the user.

10. The method of claim 1 , further comprising, prior to selecting a subset of the items on the recommendations list, augmenting the recommendations list with at least one item that, based on recorded activity of the user, was removed from an electronic shopping cart of the user without being purchased by the user.

11. An item recommendation system, comprising: physical, non-transitory computer storage that stores a mapping that maps individual items to respective sets of related items; and a recommendation component that is operative to generate personalized recommendations for a user by a process that comprises: identifying a plurality of items of interest to the user; for each item of interest, obtaining, from the mapping, a respective related items list; weighting the related items lists based on information reflective of the user's affinity for the corresponding items of interest, wherein weighting a related items list comprises weighting individual items on the related items lists, and wherein the amounts by which the items are weighted influences a likelihood that such items will be selected to recommend to the user, wherein weighting the related items lists comprises at least one of the following: (1) weighting a related items list by an amount that depends upon an amount of time that has transpired since the user performed an action that evidences the user's affinity for the corresponding item of interest, and (2) weighting a related items list by an amount that depends upon an explicit rating by the user of the corresponding item of interest; generating a recommendations pool at least partly by combining the weighted related items lists, said recommendations pool including data values for respective items in the pool, each data value being dependent upon (1) whether the respective item is on more than one of the related items lists, and (2) the weighting applied to the related items list or lists on which the respective item appears; and selecting, based at least partly on said data values, a subset of the items on the recommendations list to recommend to the user.

12. The item recommendation system of claim 11 , wherein each related items list comprises a set of relatedness values, each relatedness value corresponding to a respective item on the related items list and representing a degree to which that item is related to the corresponding item of interest, and wherein the recommendation component is operative to weight the related items lists by multiplying each of the relatedness values in a list by the weighting amount.

13. The item recommendation system of claim 11 , wherein the recommendations component is operative to weight a related items list by an amount that depends upon an amount of time that has transpired since the user performed an action that evidences the user's affinity for the corresponding item of interest.

14. The item recommendation system of claim 13 , wherein said action is a purchase by the user of the corresponding item of interest.

15. The item recommendation system of claim 11 , wherein the recommendation component is operative to weight a related items list by an amount that depends upon an explicit rating by the user of the corresponding item of interest.

16. The item recommendation system of claim 11 , further comprising an off-line component that is operative to generate the mapping based at least partly on detected correlations between item interests of users.

17. The item recommendation system of claim 11 , wherein the mapping is based at least partly on content-based relationships between particular items.

18. The item recommendation system of claim 11 , wherein the recommendations component is operative to identify the plurality of items of interest based on current contents of an electronic shopping cart of the user.

19. The item recommendation system of claim 11 , wherein the recommendations component is operative to identify the plurality of items of interest by selecting at least some of the items of interest from a recorded item viewing history of the user.

20. The item recommendation system of claim 11 , wherein the item recommendation system comprises a plurality of physical machines, each having stored thereon (1) a copy of said mapping, and (2) executable code for implementing said process.

21. Physical, non-transitory computer storage having stored thereon executable code that directs a computer system to perform a process that comprises: identifying a plurality of items that, based on activity of a user, are deemed to be of interest to the user; for each of said items of interest, retrieving, from a pre-generated data structure that maps items to related items, a respective related items list; weighting the related items lists based on information reflective of the user's affinity for the corresponding items of interest, such that at least some of the related items lists are weighted differently than others, wherein weighting a related items list comprises weighting individual items on the related items lists, and wherein the amounts by which the items are weighted influences a likelihood that such items will be selected to recommend to the user, wherein weighting the related items lists comprises at least one of: (1) weighting a related items list by an amount that depends upon an amount of time that has transpired since the user performed an action that evidences the user's affinity for the corresponding item of interest, and (2) weighting a related items list by an amount that depends upon an explicit rating by the user of the corresponding item of interest; generating a recommendations list at least partly by combining the weighted related items lists, said recommendations lists including data values for respective items thereon, each data value being dependent upon (1) whether the respective item is on more than one of the related items lists, and (2) the weighting applied to the related items list or lists on which the respective item appears; and selecting, based at least partly on said data values, a subset of the items on the recommendations list to recommend to the user.

22. The non-transitory computer storage of claim 21 , wherein the process comprises weighting a related items list by an amount that depends upon an amount of time that has transpired since the user performed an action that evidences the user's affinity for the corresponding item of interest.

23. The non-transitory computer storage of claim 21 , wherein the process comprises weighting a related items list by an amount that depends upon an explicit rating by the user of the corresponding item of interest.

24. The non-transitory computer storage of claim 21 , in combination with said computer system, wherein the computer system comprises one or more computers, and is programmed with said executable code to perform said process.